News for hackers. Stuff that matters.
Re:Show HN: A neural net that critiques your logo by neilk on Thursday August 31, 2017 @10:02AM (#15140342)
I tried our company's logo (https://bench.co/press-kit/#brand), which I think is pretty strong, and it gave it high marks.

But then I tried a very bad logo ("Kate's Florist" from http://www.paulmurraydesign.com/graphic-design/how-to-spot-a...) and it was ranked almost as good.

Re:Show HN: A neural net that critiques your logo by alexwebb2 on Thursday August 31, 2017 @03:43PM (#15143250)
I think the Bench logo is confusing. I've never heard of the company before, but the logo looks like it might read "Bench Pi". It's the pi symbol in a shield (although it's so small it's very hard to tell at first glance), so... I'd guess the company has something to do with setting math-based benchmarks as some sort of metric-as-a-service?

[looks at about page]

Okay, no, it's nothing like that at all. The logo gave me the wrong first impression of the company.

Re:Show HN: A neural net that critiques your logo by amelius on Thursday August 31, 2017 @04:42PM (#15143747)
The bench looks like a "registered/trademark" symbol (it's in the same position). That's confusing.
Re:Show HN: A neural net that critiques your logo by apk-d on Friday September 01, 2017 @05:35AM (#15147180)
It just looks like something akin to superscript to me. I did not associate it with ©®™ at all.
Re:Show HN: A neural net that critiques your logo by grey-area on Thursday August 31, 2017 @10:49AM (#15140678)
So the rules given are Uniqueness, Legibility, Color/Contrast with a very simplistic description of each which I imagine reflects the rules used. That logo passes with flying colours according to their criteria, perhaps lower marks for legibility. Needless to say, there are more things to logo design than are dreamt of in this philosophy.

There's more about how they use deep learning here:

http://brandmark.io/intro/

but for me it seems the constraints they've chosen are ill-conceived (things like matching logo strength to font weight) and unlikely to come up with something inspiring or even decent a lot of the time or rank things well which humans would have no difficulty recognising. This is a difficult problem, and I wouldn't expect rules based solutions to work very well, though some kind of permutation engine starting from known good designs could work better.

Re:Show HN: A neural net that critiques your logo by raganesh on Thursday August 31, 2017 @03:27PM (#15143128)
Good point.

I guess it is a fair assumption that the neural network is trained using a corpus of logos that already look "well made" to the human eye.

It is possible there may not have been any/many samples that would be considered bad.

It would be interesting to see how this neural net evaluates a bad logo, using a tool such as LIME.